Graph-Based Fraud Detection with Dual-Path Graph Filtering

Wei He; Wensheng Gan; Philip S. Yu

cs.LGApr 14, 2026

Graph-Based Fraud Detection with Dual-Path Graph Filtering

Authors: Wei He, Wensheng Gan, Philip S. Yu

Organizations: School of Intelligent Systems Science and Engineering, Jinan University, Zhuhai 519070, China · Department of Computer Science, University of Illinois Chicago, Chicago 60607, USA

Abstract

Fraud detection on graph data can be viewed as a demanding task that requires distinguishing between different types of nodes. Because graph neural networks (GNNs) are naturally suited for processing information encoded in graph form through their message-passing operations, methods based on GNN models have increasingly attracted attention in the fraud detection domain. However, fraud graphs inherently exhibit relation camouflage, high heterophily, and class imbalance, causing most GNNs to underperform in fraud detection tasks. To address these challenges, this paper proposes a Graph-Based Fraud Detection Model with Dual-Path Graph Filtering (DPF-GFD). DPF-GFD first applies a beta wavelet-based operator to the original graph to capture key structural patterns. It then constructs a similarity graph from distance-based node representations and applies an improved low-pass filter. The embeddings from the original and similarity graphs are fused through supervised representation learning to obtain node features, which are finally used by an ensemble tree model to assess the fraud risk of unlabeled nodes. Unlike existing single-graph smoothing approaches, DPF-GFD introduces a frequency-complementary dual-path filtering paradigm tailored for fraud detection, explicitly decoupling structural anomaly modeling and feature similarity modeling. This design enables more discriminative and stable node representations in highly heterophilous and imbalanced fraud graphs. Comprehensive experiments on four real-world financial fraud detection datasets demonstrate the effectiveness of our proposed method.

Figures & tables

Figure 1: Overview of the proposed DPF-GFD framework. The model adopts a dual-path graph filtering design to address relation camouflage, heterophily, and class imbalance in fraud graphs. The structural path enhances multi-frequency structural anomalies on the original graph, while the similarity-based path applies low-pass smoothing on a kNN graph to reconstruct feature consistency. The fused representations are then fed into an imbalance-robust ensemble classifier for fraud risk prediction.

Figure 2: Beta wavelet-based kernels in the spectral domain when

C=4

Dataset	FDCompCN	FFSD	Elliptic	DGraph
Nodes	5317	77881	203769	3700550
Edges	10059	860968	234355	4300999
Features	57	7	166	17
Normals	4758	24387	42019	1210092
Frauds	559	5256	4545	15509
Unknowns	0	48238	157205	2474949

Table 1: Statistical information of the four datasets used for evaluation, including the total number of nodes, edges, feature dimensions, normal nodes, fraudulent nodes, and unknown nodes.

Models	FDCompCN				FFSD
	Rec@K	F1	AUC	AP	Rec@K	F1	AUC	AP
GCN	0.1905 $\pm$ 0.0084	0.2245 $\pm$ 0.0202	0.6065 $\pm$ 0.0318	0.1763 $\pm$ 0.0087	0.3914 $\pm$ 0.0511	0.3763 $\pm$ 0.0487	0.6768 $\pm$ 0.0284	0.4240 $\pm$ 0.0330
GraphSAGE	0.1810 $\pm$ 0.0308	0.2185 $\pm$ 0.0257	0.6031 $\pm$ 0.0434	0.1778 $\pm$ 0.0154	0.2520 $\pm$ 0.1162	0.2907 $\pm$ 0.1712	0.5780 $\pm$ 0.1004	0.2592 $\pm$ 0.0787
GAT	0.1667 $\pm$ 0.0304	0.1936 $\pm$ 0.0140	0.5784 $\pm$ 0.0311	0.1643 $\pm$ 0.0210	0.4324 $\pm$ 0.0091	0.4662 $\pm$ 0.0119	0.6564 $\pm$ 0.0054	0.3962 $\pm$ 0.0213
GAS	0.1857 $\pm$ 0.0136	0.2369 $\pm$ 0.0128	0.6358 $\pm$ 0.0175	0.1812 $\pm$ 0.0097	0.4365 $\pm$ 0.0224	0.4587 $\pm$ 0.0159	0.7032 $\pm$ 0.0233	0.4347 $\pm$ 0.0556
PCGNN	0.3357 $\pm$ 0.0155	0.3131 $\pm$ 0.0468	0.7465 $\pm$ 0.0155	0.3050 $\pm$ 0.0189	0.3123 $\pm$ 0.1185	0.3354 $\pm$ 0.1446	0.6253 $\pm$ 0.0464	0.3456 $\pm$ 0.0988

Table 2: Performance comparison (Mean

\pm

Std). The best results are highlighted in bold.

Figure 3: Results of the ablation study on four datasets.

Models	FDCompCN	FFSD	Elliptic	DGraph
Raw + XGB	$0.4247\pm 0.0153$	$0.7497\pm 0.0070$	$0.7571\pm 0.0035$	$0.0271\pm 0.0003$
GCN Emb + XGB	$0.2163\pm 0.0196$	$0.6170\pm 0.0095$	$0.5052\pm 0.0440$	$0.0320\pm 0.0003$
DPF-GFD	$\mathbf{0.4537\pm 0.0165}$	$\mathbf{0.7679\pm 0.0025}$	$\mathbf{0.7819\pm 0.0045}$	$\mathbf{0.0347\pm 0.0006}$

Table 3: Control experiments with a fixed XGBoost classifier. Results are reported in AP (mean

\pm

std over 5 runs).

Figure 4: Performance under different values of parameter

C

on four datasets.

Figure 5: Performance under different values of parameter K on four datasets.

Figure 6: Visualization of the learned node embeddings on the FDCompCN dataset.

Explore similar work

Beyond Sparse Supervision: Diffusion-Guided Learning for Few-Shot Graph Fraud Detection
Jun 26, 2026Liming Liu, Chao Hu, Mingfei Lu +3Graph Neural Networks Graph SSL
Central South University, Changsha, China · University of Technology Sydney, Australian Artificial Intelligence Institute, Sydney, Australia
A Novel Graph Fraud Detector via Grouped Attribute Completion and Confidence-Aware Contrastive Learning
Jul 13, 2026Junpeng Wu, Ye YuanGraph Neural Networks Graph-Based Semi-Supervised Learning
College of Computer and Information Science Southwest University Chongqing, China
Multi-Scale Adaptive Neighborhood Awareness Transformer For Graph Fraud Detection
Mar 3, 2026Jiaqi Lv, Qingfeng Du, Yu Zhang +2Graph Transformer Anomaly Detection
School of Computer Science and Technology, Tongji University, Shanghai, China

Jun 26, 2026J/K move · Enter open · S save

Beyond Sparse Supervision: Diffusion-Guided Learning for Few-Shot Graph Fraud Detection

Liming Liu, Chao Hu, Mingfei Lu, Yiwei Ge+2

Central South University, Changsha, China · University of Technology Sydney, Australian Artificial Intelligence Institute, Sydney, Australia

Graph Neural Networks Graph SSL Financial Fraud Detection Few-Shot Learning Spectral Graph Convolution Anomaly Detection

Open paper AskPDF arXiv

Graph-based fraud detection is essential for safeguarding large-scale transaction systems, where undetected anomalies may lead to substantial financial losses and security risks. Real-world fraud graphs pose two coupled challenges: sparse and imbalanced supervision, where verified fraudulent labels are scarce and heavily skewed toward benign accounts, and representation dilution, where spatial message passing may oversmooth camouflaged anomalies while spectral filters may suppress fraud-relevant mid- and high-frequency irregularities. To address these challenges, we propose ADC-GNN, short for Attention-guided Diffusion-Contrastive Graph Neural Network, a unified framework that combines diffusion-guided feature augmentation, contrastive representation learning, and multi-hop spectral attention for few-shot graph fraud detection. The diffusion component is formulated as a feature-space denoising augmentation mechanism rather than a full topology-generative graph diffusion model: it constructs noise-perturbed node-feature views under a cosine schedule and uses contrastive learning to stabilize node representations across perturbations. The spectral attention module further adaptively emphasizes fraud-relevant hop-level and relation-level cues. We evaluate ADC-GNN primarily on three public benchmarks and additionally report a proprietary real-world telecom transaction dataset with approximately 60,000 records as a private case study. Under the 1% training setting, ADC-GNN achieves consistent improvements over original graph fraud baselines and four protocol-consistent recent graph anomaly/fraud baselines on the public benchmarks. Additional analyses on split stability, training ratios, oversampling alternatives, module-level ablations, diffusion schedules, and runtime and memory-consumption comparisons further characterize the effective operating regime of ADC-GNN.

Abstract

Explore similar work

Beyond Sparse Supervision: Diffusion-Guided Learning for Few-Shot Graph Fraud Detection

A Novel Graph Fraud Detector via Grouped Attribute Completion and Confidence-Aware Contrastive Learning

Multi-Scale Adaptive Neighborhood Awareness Transformer For Graph Fraud Detection